Re: Segfaults during production of documentation

[Daniel J Sebald]

On 01/07/2018 11:02 PM, Dmitri A. Sergatskov wrote:
> On Sun, Jan 7, 2018 at 10:55 PM, Dmitri A. Sergatskov 
> <address@hidden <mailto:address@hidden>> wrote:
[snip]
>     ​Scaling looks fine. I do run make as:
>
>     LIBGL_ALWAYS_SOFTWARE=1 LD_PRELOAD=/usr/lib64/libGLX_mesa.so.0
>     ​make -j8​
>
>
>     The old ​crashed due to​ bad GL required restarting the make. Now it
>     just goes through.
>
>     Dmitri.
>     -- 
>
>
> ​OK I was able to get plot windows to some funny state by shrinking it 
> first to 0 (so no plot area is visible)
> and then *quickly​* stretching it out. So it is not perfected yet...

That's a known bug:

https://savannah.gnu.org/bugs/?func=detailitem&item_id=52398

Please try again to produce a black background, in particular the FLTK 
terminal.  After creating the plot, keep dragging the lower right corner 
around for 30 s.  It seems about every seven or eight seconds there is a 
black background.

Now that I try Qt graphics toolkit again, maybe the Qt g.t. terminal is 
not flashing to a black background.  It's only the outside edges that 
appear black as the window is expanded.  But the refresh on the graphics 
routines should not be so slow that the black edges are so noticeable. 
The Qt graphics is going directly to the OpenGL routines.  gnuplot I 
could understand being slower because there is a lot going on between 
its refresh...but gnuplot is pretty responsive.

Just speculation, but I wonder if there was some graphics bug for which 
the solution was to introduce a delay in the refresh, and maybe in 
offline mode there is no delay and the bug is re-emerging?

Dan